Have you ever noticed this from KSP tutorial ex52? I'm wondering whether it could be related with PETSc interface to MUMPS or it is a problem on the MUMPS side. The valgrind error come out once the first time the KSP is solved, then disappear if KSPSolve is repeatedly called.
==31042== Command: ./ex52 -use_mumps_lu ==31042== Parent PID: 31041 ==31042== ==31042== Syscall param writev(vector[...]) points to uninitialised byte(s) ==31042== at 0x70C761B: writev (writev.c:51) ==31042== by 0x8262108: MPIDU_Sock_writev (sock_immed.i:610) ==31042== by 0x824B0DB: MPIDI_CH3_iSendv (ch3_isendv.c:84) ==31042== by 0x823F4F2: MPIDI_CH3_PktHandler_RndvClrToSend (ch3u_rndv.c:249) ==31042== by 0x824C417: MPIDI_CH3I_Progress_handle_sock_event (ch3_progress.c:453) ==31042== by 0x824C4D7: MPIDI_CH3i_Progress_test (ch3_progress.c:95) ==31042== by 0x82A9AAF: MPIR_Test_impl (test.c:50) ==31042== by 0x82A9B88: PMPI_Test (test.c:151) ==31042== by 0x8224829: mpi_test (testf.c:200) ==31042== by 0x61A8805: __dmumps_comm_buffer_MOD_dmumps_4 (dmumps_comm_buffer.F:635) ==31042== by 0x61A8D83: __dmumps_comm_buffer_MOD_dmumps_62 (dmumps_comm_buffer.F:554) ==31042== by 0x6230D6E: dmumps_242_ (dmumps_part2.F:739) ==31042== Address 0x9d0af38 is 8 bytes inside a block of size 336 alloc'd ==31042== at 0x4C2B6CD: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so) ==31042== by 0x61AAE8F: __dmumps_comm_buffer_MOD_dmumps_2 (dmumps_comm_buffer.F:175) ==31042== by 0x61AB27C: __dmumps_comm_buffer_MOD_dmumps_55 (dmumps_comm_buffer.F:123) ==31042== by 0x6186B64: dmumps_301_ (dmumps_part8.F:989) ==31042== by 0x61FBA10: dmumps_ (dmumps_part1.F:665) ==31042== by 0x60DE3FB: dmumps_f77_ (dmumps_part3.F:6651) ==31042== by 0x60B7219: dmumps_c (mumps_c.c:422) ==31042== by 0x59AE73F: MatSolve_MUMPS (mumps.c:591) ==31042== by 0x58B1DB7: MatSolve (matrix.c:3178) ==31042== by 0x5CB6A19: PCApply_LU (lu.c:198) ==31042== by 0x5C1E9DE: PCApply (precon.c:442) ==31042== by 0x5E0FE44: KSP_PCApply (kspimpl.h:227) -- Ph. D. Stefano Zampini CINECA SuperComputing Applications and Innovations Department - SCAI Via dei Tizii, 6 00185 Roma - ITALY ------------------------------------------------------------------------------------------------------------------------ Email: s.zamp...@cineca.it SkypeID: stefano.zampini GoogleTalk: stefano.zamp...@gmail.com Tel: +39 06.44486.707 ------------------------------------------------------------------------------------------------------------------------
